Applying GLF-Based Group-Level Analysis to Single-Subject Data

AFNI Message Board
,
1

I used 3dLMEr to test the effects of interest at the population level with the following command:

3dLMEr -prefix LME2_GLM_delay_bycond -jobs 12 \
    -mask /data/fmri/freesurfer/subjects/MNI152/SUMA/ribbon_mask2+tlrc \
    -model 'load+(1|Subj)' \
    -glfCode linear_effect 'load : 1*CI -0.5*SGI1 -0.5*SGI2 & 0.5*SGI1 +0.5*SGI2 -1*SGI3' \
    -dataTable @dataTable.txt

The dataTable.txt file contains the following entries:

Subj  load  InputFile
s1    CI    sub20_glm_LME_delay_cond_CI+tlrc
s1    SGI1  sub20_glm_LME_delay_cond_SGI1+tlrc
s1    SGI2  sub20_glm_LME_delay_cond_SGI2+tlrc
s1    SGI3  sub20_glm_LME_delay_cond_SGI3+tlrc
s2    CI    sub21_glm_LME_delay_cond_CI+tlrc
s2    SGI1  sub21_glm_LME_delay_cond_SGI1+tlrc
s2    SGI2  sub21_glm_LME_delay_cond_SGI2+tlrc
s2    SGI3  sub21_glm_LME_delay_cond_SGI3+tlrc
...

The input files are from the outputs from a GLM analysis performed using 3dDeconvolve with the following parameters:

3dDeconvolve -input pb01.imagery.r??.volreg+tlrc.HEAD  \
                  -censor txt/censor_imagery_combined_2.1D   \
	          -mask /data/fmri/freesurfer/subjects/MNI152/SUMA/ribbon_mask2+tlrc \
                  -num_stimts 12 \
                  -stim_times 1 txt/Mixed_sample.txt "BLOCK4(1.6,1)" -stim_label 1 "Sample" \
                  -stim_times 2 txt/CI_delay_onset.txt "BLOCK4(5.6,1)" -stim_label 2 "DelayCI" \
                  -stim_times 3 txt/SGI1_delay_onset.txt "BLOCK4(5.6,1)" -stim_label 3 "DelaySGI1" \
                  -stim_times 4 txt/SGI2_delay_onset.txt "BLOCK4(5.6,1)" -stim_label 4 "DelaySGI2" \
                  -stim_times 5 txt/SGI3_delay_onset.txt "BLOCK4(5.6,1)" -stim_label 5 "DelaySGI3" \
                  -stim_times 6 txt/Mixed_probe.txt "BLOCK4(2,1)" -stim_label 6 "Probe" \
                  -stim_file 7 txt/motion_demean.1D'[0]' -stim_label 7 "Roll" -stim_base 7 \
                  -stim_file 8 txt/motion_demean.1D'[1]' -stim_label 8 "Pitch" -stim_base 8 \
                  -stim_file 9 txt/motion_demean.1D'[2]' -stim_label 9 "Yaw" -stim_base 9 \
                  -stim_file 10 txt/motion_demean.1D'[3]' -stim_label 10 "dS" -stim_base 10 \
                  -stim_file 11 txt/motion_demean.1D'[4]' -stim_label 11 "dL" -stim_base 11 \
                  -stim_file 12 txt/motion_demean.1D'[5]' -stim_label 12 "dP" -stim_base 12 \
                  -tout -nofull_first -polort 3 -bucket tlrc_GLM_bycond/glm_LME_delay_cond -jobs 24

Now, I would like to conduct the same analysis specified by -glfCode but at the single-subject level in the original space. Should I use -gltSym while performing 3dDeconvolve for individual subjects, or would it be better to perform an additional analysis using 3dANOVA? If so, could you provide guidance on how to approach this for each method?

2

Try adding the following to your 3dDeconvolve script:

-gltsym '+DelayCI -0.5*DelaySGI1 -0.5*DelaySGI2 \ +0.5*DelaySGI1 +0.5*DelaySGI2 -DelaySGI3'

Gang Chen

3

Thank you for your reply!
Shitong Zhao